大模型推理能力

Case-Based Reasoning with Language Models for Classification of Logical Fallacies

Zhivar Sourati, Filip Ilievski, Hông-Ân Sandlin, Alain Mermoud

专题命中 复杂问题求解 ：reasoning(title,abstract)；分类 cs.CL、cs.AI

Graph-ToolFormer: To Empower LLMs with Graph Reasoning Ability via Prompt Augmented by ChatGPT

Jiawei Zhang

专题命中 复杂问题求解 ：reasoning(title,abstract)；分类 cs.AI、cs.LG

Comments 34 pages, 3 figures, 8 tables

Explainable Verbal Reasoner Plus (EVR+): A Natural Language Reasoning Framework that Supports Diverse Compositional Reasoning

Zhengzhong Liang, Zeyu Zhang, Steven Bethard, Mihai Surdeanu

专题命中 复杂问题求解 ：reasoning(title,abstract)；分类 cs.CL、cs.AI

Iterated Decomposition: Improving Science Q&A by Supervising Reasoning Processes

Justin Reppert, Ben Rachbach, Charlie George, Luke Stebbing, Jungwon Byun, Maggie Appleton, Andreas Stuhlmüller

专题命中 复杂问题求解 ：reasoning(title,abstract)；分类 cs.CL、cs.AI

Visual Programming: Compositional visual reasoning without training

Tanmay Gupta, Aniruddha Kembhavi

专题命中 复杂问题求解 ：reasoning(title,abstract)；分类 cs.CL、cs.AI

Teaching Broad Reasoning Skills for Multi-Step QA by Generating Hard Contexts

Harsh Trivedi, Niranjan Balasubramanian, Tushar Khot, Ashish Sabharwal

专题命中 复杂问题求解 ：reasoning(title,abstract)；分类 cs.CL、cs.AI

Comments Accepted at EMNLP'22

Great Truths are Always Simple: A Rather Simple Knowledge Encoder for Enhancing the Commonsense Reasoning Capacity of Pre-Trained Models

Jinhao Jiang, Kun Zhou, Wayne Xin Zhao, Ji-Rong Wen

专题命中 复杂问题求解 ：reasoning(title,abstract)；分类 cs.CL、cs.AI

Comments 12 pages, NAACL-Findings-2022

ComPhy: Compositional Physical Reasoning of Objects and Events from Videos

Zhenfang Chen, Kexin Yi, Yunzhu Li, Mingyu Ding, Antonio Torralba, Joshua B. Tenenbaum, Chuang Gan

专题命中 复杂问题求解 ：reasoning(title,abstract)；分类 cs.AI、cs.LG

Comments ICLR 2022. Project page: https://comphyreasoning.github.io/

JointLK: Joint Reasoning with Language Models and Knowledge Graphs for Commonsense Question Answering

Yueqing Sun, Qi Shi, Le Qi, Yu Zhang

专题命中 复杂问题求解 ：reasoning(title,abstract)；分类 cs.CL、cs.AI

Comments Accepted by NAACL 2022 main conference (Long paper)

Dynamic Key-value Memory Enhanced Multi-step Graph Reasoning for Knowledge-based Visual Question Answering

Mingxiao Li, Marie-Francine Moens

专题命中 复杂问题求解 ：reasoning(title,abstract)；分类 cs.CL、cs.AI

Recursive Reasoning Graph for Multi-Agent Reinforcement Learning

Xiaobai Ma, David Isele, Jayesh K. Gupta, Kikuo Fujimura, Mykel J. Kochenderfer

专题命中 复杂问题求解 ：reasoning(title,abstract)；分类 cs.AI、cs.LG

Comments AAAI 2022

GreaseLM: Graph REASoning Enhanced Language Models for Question Answering

Xikun Zhang, Antoine Bosselut, Michihiro Yasunaga, Hongyu Ren, Percy Liang, Christopher D. Manning, Jure Leskovec

专题命中 复杂问题求解 ：reasoning(title,abstract)；分类 cs.CL、cs.LG

Comments Published at ICLR 2022. All code, data, and pretrained models are available at https://github.com/snap-stanford/GreaseLM

Analyzing the Effects of Reasoning Types on Cross-Lingual Transfer Performance

Karthikeyan K, Aalok Sathe, Somak Aditya, Monojit Choudhury

专题命中 复杂问题求解 ：reasoning(title,abstract)；分类 cs.CL、cs.AI

Comments Workshop on Multilingual Representation Learning (MRL 2021), at Empirical Methods in Natural Language Processing (EMNLP 2021)

CausalCity: Complex Simulations with Agency for Causal Discovery and Reasoning

Daniel McDuff, Yale Song, Jiyoung Lee, Vibhav Vineet, Sai Vemprala, Nicholas Gyde, Hadi Salman, Shuang Ma, Kwanghoon Sohn, Ashish Kapoor

专题命中 复杂问题求解 ：reasoning(title,abstract)；分类 cs.AI、cs.LG

KVL-BERT: Knowledge Enhanced Visual-and-Linguistic BERT for Visual Commonsense Reasoning

Dandan Song, Siyi Ma, Zhanchen Sun, Sicheng Yang, Lejian Liao

专题命中 复杂问题求解 ：reasoning(title,abstract)；分类 cs.CL、cs.AI

R2-B2: Recursive Reasoning-Based Bayesian Optimization for No-Regret Learning in Games

Zhongxiang Dai, Yizhou Chen, Kian Hsiang Low, Patrick Jaillet, Teck-Hua Ho

专题命中 复杂问题求解 ：reasoning(title,abstract)；分类 cs.AI、cs.LG

Comments Accepted to 37th International Conference on Machine Learning (ICML 2020), Extended version with proofs and additional experimental details and results, 27 pages

Analysing Neural Language Models: Contextual Decomposition Reveals Default Reasoning in Number and Gender Assignment

Jaap Jumelet, Willem Zuidema, Dieuwke Hupkes

专题命中 复杂问题求解 ：reasoning(title,abstract)；分类 cs.CL、cs.AI

Comments To appear at CoNLL2019

Reasoning-Driven Question-Answering for Natural Language Understanding

Daniel Khashabi

专题命中 复杂问题求解 ：reasoning(title,abstract)；分类 cs.CL、cs.AI

Comments PhD Dissertation; Presented to Computer and Information Sciences department, at the University of Pennsylvania

CLEAR: A Dataset for Compositional Language and Elementary Acoustic Reasoning

Jerome Abdelnour, Giampiero Salvi, Jean Rouat

专题命中 复杂问题求解 ：reasoning(title,abstract)；分类 cs.CL、cs.LG

Comments NeurIPS 2018 Visually Grounded Interaction and Language (ViGIL) Workshop

Compositional Language Understanding with Text-based Relational Reasoning

Koustuv Sinha, Shagun Sodhani, William L. Hamilton, Joelle Pineau

专题命中 复杂问题求解 ：reasoning(title,abstract)；分类 cs.CL、cs.AI

Comments 4 pages of main content, to be presented at Relational Representation Learning Workshop, NIPS 2018, Montreal

Pruning via Causal Attribution Preserves Reasoning Performance in Large Language Models

基于因果归因的剪枝保留大型语言模型的推理性能

Amogh Sheth, Biruk Assefa, Yi Wen Huang, Andrew Lin, Yuhao Ge

机构 * Edison Academy Magnet School（爱迪生学院磁石学校） ； Massachusetts Institute of Technology（麻省理工学院） ； State University of New York College at Plattsburgh（纽约州立大学普拉茨堡学院） ； The University of Texas at Austin（德克萨斯大学奥斯汀分校） ； Independent Researcher（独立研究员） 纠错

专题命中 复杂问题求解 ：reasoning(title,abstract)；分类 cs.CL

AI总结 提出无需训练的因果归因剪枝（CAP）方法，通过测量注意力头对推理任务的因果影响进行细粒度剪枝，在20%稀疏度下相比Wanda在ARC-Challenge上准确率提升高达61%。

Comments Accepted at the ICLR 2026 Workshop on LLM Reasoning. 13 pages, 2 figures

Decomposing LLM Self-Correction: The Accuracy-Correction Paradox and Error Depth Hypothesis

分解大语言模型的自我纠正：准确性-纠正悖论与错误深度假说

Yin Li

机构 * University of Birmingham（伯明翰大学） 纠错

专题命中 复杂问题求解 ：self-correction(title,abstract)；分类 cs.AI

AI总结 本研究揭示大语言模型在自我纠正中的准确性-纠正悖论，提出错误深度假说，发现更强模型犯更深入的错误，且错误检测与纠正成功率无直接关联。

Comments 9 pages, 2 figures, 3 tables. Code available at https://github.com/Kevin0304-li/llm-self-correction

Thinking Fast and Laterally: Multi-Agentic Approach for Reasoning about Uncertain Emerging Events

Stefan Dernbach, Alejandro Michel, Khushbu Agarwal, Christopher Brissette, Geetika Gupta, Sutanay Choudhury

专题命中 复杂问题求解 ：reasoning(title,abstract)；分类 cs.AI

Comments Presented in The 1st Workshop on System-2 Reasoning at Scale (NeurIPS 2024), Vancouver, Canada

World Programs for Model-Based Learning and Planning in Compositional State and Action Spaces

Marwin H. S. Segler

专题命中 复杂问题求解 ：planning(title,abstract)；分类 cs.LG；reasoning(comments)

Comments Accepted at the Generative Modeling and Model-Based Reasoning for Robotics and AI workshop at ICML 2019. Presented on June 14th 2019. See https://sites.google.com/view/mbrl-icml2019

Journal ref https://sites.google.com/view/mbrl-icml2019

MoT: Modularization-of-Thought Prompting for Effective Code Generation

MoT：用于有效代码生成的思维模块化提示

Ruwei Pan, Hongyu Zhang

专题命中 复杂问题求解 ：CoT(abstract,abstract_cn)；reasoning(abstract)；chain-of-thought(abstract)

AI总结 研究针对大语言模型代码生成中现有提示技术不足，提出MoT技术，利用模块化原则分解问题，用MLR图构建推理过程，经实验对比六种基线技术，在八个基准上显著提升代码生成性能，Pass@1分数达58.1%至95.1%。

Reading Between the Dots: Decoding Hidden Computation across Filler Tokens

解读点之间的信息：解码跨填充令牌的隐藏计算

Kaley Brauer, Claudio Mayrink Verdun, Samuel Marks

机构 * Harvard University（哈佛大学） ； Cambridge Boston Alignment Initiative（剑桥波士顿对齐计划） ； Massachusetts Institute of Technology（麻省理工学院） ； Anthropic 纠错

专题命中 复杂问题求解 ：reasoning(abstract)；chain-of-thought(abstract)；CoT(abstract)；分类 cs.CL、cs.AI、cs.LG

AI总结 研究前沿语言模型对无内容填充令牌的隐藏计算，通过分析两个前沿模型在四个任务家族中的表现，介绍无监督解码管道，能从隐藏状态恢复中间值，证明隐藏计算可从残差流读取。

Comments Accepted to ICML 2026 Mech Interp Workshop, 10 main paper pages, 20 appendix pages

PestVL-Net: Enabling Multimodal Pest Learning via Fine-grained Vision-Language Interaction

PestVL-Net: 通过细粒度视觉-语言交互实现多模态害虫学习

Xueheng Li, Tao Hu, Ke Cao, Runsheng Qi, Huixin Zhang, Rui Li, Jie Zhang, Chengjun Xie

机构 * Institute of Intelligent Machines, Hefei Institutes of Physical Science, Chinese Academy of Sciences（智能机器研究所，合肥物理科学研究院，中国科学院） ； University of Science and Technology of China（中国科学技术大学） ； Zhongke Hefei Institute of Technology Innovation Engineering（中科合肥技术创新工程研究院） 纠错

专题命中 复杂问题求解 ：CoT(abstract,abstract_cn)；reasoning(abstract)；chain-of-thought(abstract)

AI总结 本文提出PestVL-Net框架，结合视觉与语言模型，解决细粒度害虫识别难题，通过RWKV架构和多模态大语言模型实现高效害虫学习。

Comments 10 pages, 7 figures

Explore Briefly, Then Decide: Mitigating LLM Overthinking via Cumulative Entropy Regulation

briefly, Then Decide: 通过累积熵调节缓解LLM过度思考

Yi Bin, Tianyi Jiang, Yujuan Ding, Kainian Zhu, Fei Ma, Jingkuan Song, Yang Yang, Heng Tao Shen

机构 * Tongji University（同济大学） ； Hong Kong Polytechnic University（香港理工大学） ； Shanghai University of Electric and Power（上海电力大学） ； Guangdong Laboratory of Artificial Intelligence and Digital Economy（广东省人工智能与数字经济实验室） ； University of Electronic Science and Technology of China（电子科技大学） 纠错

专题命中 复杂问题求解 ：reasoning(abstract)；chain-of-thought(abstract)；CoT(abstract)；分类 cs.CL、cs.AI、cs.LG

AI总结 本文提出TECA指标和

Comments Code: https://github.com/AusertDream/CumulativeEntropyRegulation

LLM Probability Concentration: How Alignment Shrinks the Generative Horizon

LLM概率集中：对齐如何缩小生成范围

Chenghao Yang, Sida Li, Ari Holtzman

专题命中 复杂问题求解 ：reasoning(abstract)；chain-of-thought(abstract)；CoT(abstract)；分类 cs.CL、cs.AI、cs.LG

AI总结 研究发现对齐微调通过减少生成多样性，使LLM生成更一致，从而影响复杂推理稳定性。

Comments Codebase: https://github.com/yangalan123/LLMBranchingFactor. V3: Significantly rewrite the whole paper for a clearer structure. Correct problems in the theory parts (Remove emphasis on AEP, discussions on variable LLM generation lengths) and strengthen asymptotic analysis. Add Qwen and OLMo2 experiments. Preliminary SFT v.s. RL comparison to better understand the alignment effects on BF

A Vision-Language-Action Model with Visual Prompt for OFF-Road Autonomous Driving

一种具有视觉提示的视觉-语言-动作模型用于越野自动驾驶

Liangdong Zhang, Yiming Nie, Haoyang Li, Fanjie Kong, Baobao Zhang, Shunxin Huang, Kai Fu, Chen Min, Liang Xiao

专题命中 复杂问题求解 ：reasoning(abstract)；chain-of-thought(abstract)；CoT(abstract)；planning(abstract)

AI总结 本文提出OFF-EMMA模型，通过视觉提示和COT-SC策略提升越野自动驾驶轨迹规划的准确性和鲁棒性。